Overview
Anonymising a contributor is a technical problem where the failure mode is severe and delayed: a source is identified weeks later from a detail nobody considered, and the consequences fall entirely on them. Several widely used techniques provide much less protection than they appear to. This guide covers what genuinely works for picture and voice, the identifying details that survive obvious anonymisation, and the file-level traces that give people away after publication.
What You Need
- An honest conversation with the source about what you can and cannot guarantee
- A location that does not itself identify them
- Editing tools capable of proper voice alteration, not just pitch shifting
- A metadata stripping step before publication
- Someone other than you to review the cut for identifying details
- A retention decision: how long you keep the unedited material, and where
Steps
Establish what level of protection is actually needed
Protecting someone from casual recognition by a general audience is a different problem from protecting them from an employer, or from a state actor with resources. Ask who the source is worried about, and be honest that stronger adversaries can defeat weaker techniques. The answer determines everything that follows.
Prefer techniques that remove information over ones that obscure it
Silhouetting against a bright background, filming hands or an object instead of a face, shooting extreme close-ups of a detail, or reconstructing with an actor all remove identifying information entirely. Blurring and pixelation obscure information that is still present in the frame, and both have been reversed in some circumstances. Removal beats obscuring.
Treat voice as seriously as face
Voices are recognisable to anyone who knows the person, and simple pitch shifting is reversible and often defeats the purpose while sounding obviously altered. Where anonymity genuinely matters, the robust options are re-voicing the interview with another person reading the transcript, or subtitling with no audio at all.
Audit the frame for everything else identifying
A silhouette is useless if the background is their office, the shot includes a distinctive tattoo, a lanyard, a specific vehicle, a window view, or a recognisable piece of furniture. Have someone who was not at the shoot review the cut specifically for identifying detail. The person who filmed it is the least likely to notice.
Check what the words themselves reveal
A source describing an incident only four people witnessed has identified themselves to those four regardless of how the picture is treated. Review the transcript for details that narrow the field (job titles, dates, specific events, turns of phrase) and generalise or cut them in consultation with the source.
Strip metadata and control the unedited material
Published files can carry location coordinates, device identifiers, and timestamps. Strip them deliberately before publication. Then decide what happens to the original recording: where it is stored, who can access it, whether it is encrypted, and whether it should be destroyed, because material you hold can be compelled or compromised.
Pro Tips
- Never promise absolute anonymity. Explain what you will do and what its limits are, and let the source decide.
- Blurring and pixelation are weaker than they look. Prefer removing information to obscuring it.
- Pitch-shifted voice is both recognisable and reversible. Re-voicing with another reader is the robust option.
- Have a fresh pair of eyes review the cut for identifying detail. Familiarity blinds the person who shot it.
- Decide the retention policy for raw material before you record, not after someone asks for it.
Knowledge Base
What You'll Learn
Source protection fails in specific, repeatable ways. Below: why some standard techniques provide less protection than assumed, and the categories of identifying information people forget.
What Works, and What Only Looks Like It Works
Anonymisation techniques differ substantially in robustness, and the visually obvious ones are not the strongest.
Blurring and pixelation obscure information that remains present in the underlying image. Depending on how they were applied and what else is available, they have been defeated, particularly where the same subject appears across multiple frames or where the obscured region is small relative to the frame. They are adequate against casual recognition and should not be relied on against a determined adversary.
Silhouetting is stronger because it genuinely removes information. A properly exposed silhouette against a bright background contains no facial detail to recover. It fails when exposure is imperfect and detail remains visible in the shadow, so check the result carefully rather than assuming.
Not filming the person (using hands, objects, location, or reconstruction) is strongest of all, because there is nothing to recover.
Pitch shifting is the weakest common technique. It is frequently reversible, and even where it is not, prosody and speech patterns remain recognisable to people who know the speaker.
The Identifying Information People Forget
Most anonymisation failures are not defeats of the technique. They are details nobody considered as identifying.
Context. The background, the visible weather, a view from a window, a vehicle, a uniform, or a lanyard. Anything that narrows location or employer narrows the pool of possible people, sometimes to one.
Body and mannerism. Build, posture, a tattoo, jewellery, hands, and characteristic gestures are all recognisable to people who know the subject, which is exactly the group that matters.
The account itself. This is the most commonly missed. A source describing something only a handful of people knew has identified themselves to those people no matter how the footage is treated. This requires editorial handling, generalising detail, rather than technical handling.
File metadata. Published media can carry GPS coordinates, device identifiers, and precise timestamps. Stripping it must be a deliberate step, because several tools preserve it and several platforms do not strip everything.
The reliable process is a review by someone who was not present, looking specifically for these four categories.
Where This Fits
This guide covers one specific part of mobile journalism. The wider picture (shooting, filing, verifying, and the practical kit for reporting alone from the field) is in Mobile Journalism: A Complete Phone-Based Reporting Kit, which frames the discipline as a whole and links out to the detailed guides underneath it, including this one. If you are starting from scratch rather than solving a specific problem, read that first and come back here.
FAQ
Q: Is blurring a face enough to protect a source?
A: Against casual recognition, usually. Against a determined adversary, not reliably: blurring obscures information that is still present, and has been defeated in some circumstances. Where the stakes are high, prefer techniques that remove information entirely: silhouette, filming hands or objects, or not filming the person at all.
Q: How should I disguise a voice?
A: Not with pitch shifting, which is often reversible and leaves prosody and speech patterns recognisable to anyone who knows the person. The robust approaches are re-voicing the interview with a different reader working from the transcript, or presenting it as subtitles with no original audio.
Q: What identifies a source besides their face and voice?
A: Context that narrows location or employer (background, window views, uniforms, vehicles, lanyards) plus build, tattoos, jewellery, hands, and mannerisms. And most commonly missed: the account itself, since describing something only a few people knew identifies the source to exactly those people regardless of the picture treatment.
Q: What should I do with the unedited recording?
A: Decide before you record. Where it is stored, who can access it, whether it is encrypted, and whether it should be destroyed after publication. Material you continue to hold can be compelled or compromised, so retention is part of source protection rather than an administrative afterthought.
Translate this page
- Español
- 简体中文
- हिन्दी
- العربية
- Português
- Français
- Deutsch
- 日本語
- Русский
- Bahasa Indonesia
- 한국어
- Italiano
- Türkçe
- Tiếng Việt
- Polski
- Nederlands
Machine translation provided by Google Translate, on Google’s servers. We do not check these translations and they will get technical terms wrong. The English page is the authoritative one. Following a link sends this page’s address to Google. Your browser may also offer to translate this page itself, which keeps the request on your device.